lucag97/colab-server repository overview

This image allows users to open a locally running jupyter server, which allows to use the container as a local runtime in Google Colab. Normally you would run into problems when you execute some commands in Google Colab due to your registration as admin. I created a new user who has almost all rights, but is not registered as admin, which allows every command to be run from Colab.

To install the image, just use the "docker pull lucag97/colab-server" command listed above. To run an interactive container while also using the local GPU from the command line copy the following into there and adjust it to your needs: "docker run --gpus all -it --name YourName -p 8888:8888 -v "YourPath:/tf" lucag97/colab-server".

This will create a container which you can reach over port 8888 and that has mounted the directory you present in "YourPath" so you can use all those files in Colab directly. It also allows for full GPU support and only requires copying the following part from the console into Google Colab "Connect to local runtime": http://127.0.0.1:8888/tree?token=86c393fe42b14f9de61519c042cf5c6a55d20ee36335dbb8⁠ (this will look slightly different for you, since the token will be generated new everytime. You can also connect to the server via a browser using the same link.